Hybrid Quantum Error-Corrected Hadamard Edge Detection dengan Adaptive State-Vector Mean Thresholding untuk Segmentasi Citra MRI

Rakalangi, Dhanar Agastya (2026) Hybrid Quantum Error-Corrected Hadamard Edge Detection dengan Adaptive State-Vector Mean Thresholding untuk Segmentasi Citra MRI. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002221075-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (16MB) | Request a copy

Abstract

Segmentasi tumor otak pada citra Magnetic Resonance Imaging (MRI) diperlukan untuk memisahkan area tumor dari jaringan sehat sehingga lokasi dan bentuk lesi dapat dikenali secara lebih objektif. Namun, proses tersebut masih menghadapi kendala karena batas lesi sering memiliki kontras rendah, bentuk yang beragam, serta intensitas yang menyerupai jaringan di sekitarnya sehingga piksel tumor dan non-tumor sulit dibedakan. Permasalahan ini mendorong pemanfaatan informasi tepi sebagai fitur pendukung segmentasi pada citra MRI glioma, meningioma, dan pituitary. Untuk menghasilkan informasi tepi tersebut, digunakan Hybrid Quantum Error-Corrected Hadamard Edge Detection with Adaptive State-Vector Mean Thresholding (HQEHED-AMT), sedangkan klasifikasi piksel dilakukan menggunakan Particle Swarm Optimization-Support Vector Machine (PSO-SVM). HQEHED-AMT diterapkan melalui pengodean citra menggunakan Quantum Probability Image Encoding (QPIE), pemindaian horizontal dan vertikal, proses Adaptive Mean Thresholding (AMT), koreksi respons menggunakan Probabilistic Post-Quantum Error Correction (PPQEC), serta penggabungan hasil dari kedua arah pemindaian. Edge map yang dihasilkan digunakan sebagai salah satu dari sepuluh fitur piksel bersama fitur intensitas dan karakteristik lokal, sedangkan PSO digunakan untuk mengoptimalkan parameter C dan γsvm pada model SVM. Kinerja deteksi tepi dibandingkan dengan metode dasar QHED, sedangkan hasil segmentasi dibandingkan dengan metode Sobel, Prewitt, dan Laplacian of Gaussian (LoG) yang masing-masing dikombinasikan dengan PSO-SVM. Hasil pengujian menunjukkan bahwa HQEHED-AMT memperoleh nilai Figure of Merit (FOM) yang lebih tinggi daripada QHED pada ketiga jenis tumor, meskipun nilai rata-ratanya masih lebih rendah dibandingkan beberapa metode deteksi tepi klasik. Kombinasi HQEHED-AMT dan PSO-SVM menghasilkan rata-rata performa raw mask dengan nilai Dice sebesar 0,473, IoU sebesar 0,318, dan akurasi sebesar 0,952. Setelah dilakukan post-processing, performanya meningkat menjadi nilai Dice sebesar 0,600, IoU sebesar 0,441, dan akurasi sebesar 0,970. Hasil tersebut menunjukkan bahwa informasi tepi yang dihasilkan oleh HQEHED-AMT dapat mendukung segmentasi tumor otak secara efektif ketika dikombinasikan dengan fitur intensitas dan karakteristik lokal piksel.
=========================================================================================================================================
Brain tumor segmentation in Magnetic Resonance Imaging (MRI) is essential for separating tumor regions from healthy tissues, enabling the location and shape of lesions to be identified more objectively. However, this process remains challenging because lesion boundaries often exhibit low contrast, diverse shapes, and intensity characteristics similar to the surrounding tissues, making it difficult to distinguish tumor from non-tumor pixels. This challenge motivates the use of edge information as a supporting feature for the segmentation of glioma, meningioma, and pituitary MRI images. To generate this edge information, Hybrid Quantum Error-Corrected Hadamard Edge Detection with Adaptive State-Vector Mean Thresholding (HQEHED-AMT) is employed, while pixel classification is performed using Particle Swarm Optimization-Support Vector Machine (PSO-SVM). HQEHED-AMT is implemented through image encoding using Quantum Probability Image Encoding (QPIE), horizontal and vertical scanning, the Adaptive Mean Thresholding (AMT) process, response correction using Probabilistic Post-Quantum Error Correction (PPQEC), and fusion of the results from both scanning directions. The resulting edge map is used as one of ten pixel features together with intensity and local characteristic features, while PSO is employed to optimize the SVM parameters C and γsvm. Edge detection performance is compared with the baseline QHED method, whereas segmentation performance is compared with Sobel, Prewitt, and Laplacian of Gaussian (LoG), each combined with PSO-SVM. The experimental results show that HQEHED-AMT achieves a higher Figure of Merit (FOM) than the baseline QHED method for all three tumor types, although its average FOM remains lower than that of several classical edge detection methods. The combination of HQEHED-AMT and PSO-SVM produces an average raw mask performance with a Dice score of 0.473, an Intersection over Union (IoU) of 0.318, and an accuracy of 0.952. After post-processing, the performance improves to a Dice score of 0.600, an IoU of 0.441, and an accuracy of 0.970. These results indicate that the edge information generated by HQEHED-AMT can effectively support brain tumor segmentation when combined with intensity and local pixel characteristic features.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deteksi Tepi Kuantum, HQEHED–AMT, Segmentasi Tumor Otak, MRI, PSO–SVM, Quantum Edge Detection, Brain Tumor Segmentation
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > Q Science (General) > Q337.3 Swarm intelligence R Medicine > RC Internal medicine > RC78.7.N83 Magnetic resonance imaging. T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Dhanar Agastya Rakalangi
Date Deposited:	30 Jul 2026 06:19
Last Modified:	30 Jul 2026 06:19
URI:	http://repository.its.ac.id/id/eprint/139286

Actions (login required)

View Item